How To Use AI In Photoshop: Complete Guide for 2026

Last updated: October 31, 2026

Artificial intelligence has changed how artists, designers, and photographers approach image editing, turning tasks that used to take careful manual work into a few clicks and a prompt. If you’re wondering how to use AI in Photoshop, this guide walks through the main AI features Adobe documents for Photoshop on desktop, with the exact steps Adobe gives and links to Adobe’s own help pages so you can check the details yourself.

We’ll cover Generative Fill and Generative Expand, Neural Filters, and text-to-image generation, then put them together in a practical workflow. By the end, you should know where each tool lives, what it’s good for, and what to watch out for.

How AI Fits Into Photoshop Today

Photoshop’s generative features are powered by Adobe Firefly, Adobe’s family of generative AI models. Generative Fill and Generative Expand are both Firefly-powered features, and they draw on the generative credits included with Creative Cloud plans (Adobe: Generative credits overview).

Alongside the generative tools, Photoshop includes Neural Filters, a separate set of machine-learning filters for portrait edits, colorizing, style transfer and compositing. In practice, the AI tools help you extend a background, remove an unwanted object, add new elements from a text prompt, or adjust a portrait, while you keep control over the final result.

Adobe Firefly Integration

Firefly is the engine behind Photoshop’s generative features. Adobe states that its current Firefly models were trained on licensed content, such as Adobe Stock, along with public domain content, as part of its effort to design Firefly to be commercially safe (Adobe Firefly FAQ). Adobe also says it does not train Firefly on Creative Cloud subscribers’ personal content. Check Adobe’s current terms before relying on this for client work.

Cloud Processing and Connectivity

Firefly-powered features need an internet connection: Adobe lists a stable internet connection, an Adobe ID and an active entitlement as requirements for using Firefly features in Creative Cloud apps (Adobe Firefly FAQ). Some Neural Filters also use cloud processing, according to Adobe’s Neural Filters gallery. If you edit on the road, plan for that.

Mastering Generative Fill and Expand for Flawless Edits

Generative Fill and Generative Expand are the two AI features most people reach for first. They replace much of the cloning, patching, and stitching that complex image edits used to need.

Generative Fill lets you select an area of your image and fill it with new content from a text prompt, or remove what’s there. Adobe’s steps are: use a selection tool (for example the Selection Brush) to define the area, select Generative Fill, choose a model from the model picker in the Contextual Task Bar, enter a prompt or leave it blank to fill the selection from the surrounding pixels, then select Generate and choose a variation in the Properties panel (Adobe: Edit images with Generative Fill). Adobe’s model picker currently lists Adobe models (Firefly Fill & Expand, Firefly Image 5, Firefly Image 1) and partner models such as FLUX and Gemini.

Generative Expand extends your image beyond its original borders. It’s useful when a photo is cropped too tightly, or when you need more room for text or a different aspect ratio. Adobe’s route uses the Crop tool rather than Canvas Size: select the Crop tool in the Tools panel and drag the crop handles outward to expand the canvas, select Generative Expand in the Contextual Task Bar, choose a Firefly model in the model picker (Firefly Fill & Expand, Firefly Image 3 or Firefly Image 1), then leave the prompt blank or describe what you want and select Generate. Pick your favourite variation in the Properties panel (Adobe: Explore beyond the canvas with Generative Expand).

The Art of Prompting for Generative Fill

Pro tip: Generative Fill works best with clear prompts. Be specific, but not overly verbose. Instead of “add flowers,” try “add a bouquet of vibrant red roses on the table.” Mention lighting (“soft morning light”), style (“photorealistic,” “oil painting”), and surrounding elements if relevant. Experiment with different prompt lengths and descriptors, and treat it as an iterative process. For more on this, see our guide to writing better AI image prompts.

Seamless Blending and Refinement

Adobe describes Generative Fill as a non-destructive way to alter an image, and each generation gives you several variations to choose from in the Properties panel. If the first result isn’t perfect, try another variation, adjust the prompt and generate again, or use layer masks and opacity to blend the generated content with the original. Generating a few options and combining the best parts with masks keeps the AI as a co-creator rather than a replacement for your artistic eye.

Unlocking Creativity with Neural Filters

Neural Filters are AI-powered filters that use machine learning to apply creative and corrective effects. You open them from Filter > Neural Filters. Adobe splits them into Featured filters, which it describes as stable and production-ready, and Beta filters, which are still under development and whose results may vary (Adobe: Neural Filters gallery).

According to Adobe’s gallery, the Featured filters are Smart Portrait, Skin Smoothing, Super Zoom, JPEG Artifacts Removal, Colorize, Style Transfer and Makeup Transfer. The Beta filters are Photo Restoration, Harmonization, Landscape Mixer, Depth Blur and Color Transfer. Smart Portrait, for example, adjusts facial expressions, age, gaze direction, hair thickness, head direction and lighting on portraits. For composites, the Harmonization filter (Beta) matches the color and tone between layers so a subject sits more naturally in a new background.

Neural Filters Worth Knowing

These filters cover a wide range of everyday tasks (descriptions based on Adobe’s gallery):

• Smart Portrait (Featured): Adjusts facial expressions, age, gaze, head direction and lighting on portraits.
• Colorize (Featured): Adds color to black-and-white photos, with optional focal points to refine colors.
• Style Transfer (Featured): Applies an artistic style from a preset or a custom reference image.
• Landscape Mixer (Beta): Blends landscape images by combining an input image with a reference image or preset.
• Harmonization (Beta): Matches color and tone between layers for more seamless composites.

Iterative Application and Masking

Neural Filters let you choose how the result is output: as a new layer, a new masked layer, a Smart Filter, or a new document (Adobe: Use Neural Filters to enhance images). Choosing a new layer or Smart Filter keeps the edit flexible. Start with a subtle adjustment, then refine. If a filter overdoes it in one area, paint the effect back with a mask. Pro tip: combine filters for more complex effects. For example, use Smart Portrait to adjust a face, then Style Transfer to give the whole image an artistic look, refining each with its own mask.

Text-to-Image and Object Generation within Photoshop

Besides adding elements with Generative Fill, Photoshop can generate an image from a descriptive text prompt (Adobe documents this as “Generate an image with descriptive text prompts” in its Photoshop help). It’s useful for concept artists, illustrators, and anyone who needs to prototype visual ideas quickly.

This is handy for creating assets that don’t exist in your current image, or for starting a new composition from a blank canvas. Imagine needing a “steampunk airship docking at a futuristic cityscape” as a background element: type the prompt, review the variations, and keep the one that fits.

Crafting Effective Text Prompts

The quality of generated content depends heavily on your prompt. A few guidelines:

• Be Descriptive: Include details about the subject, style, colors, lighting, and composition.
• Use Keywords: “Photorealistic,” “cinematic,” “3D render,” “oil painting,” “vector art.”
• Use References: Adobe’s Photoshop help covers using reference images for more consistent results.
• Iterate and Refine: Start broad, then add details. If the first attempt isn’t right, tweak the prompt.

Integrating Generated Assets into Your Workflow

Once you’ve generated an image or object, the next step is integration:

• Scale and Position: Adjust size and placement to fit your composition. Converting the layer to a Smart Object first (Layer > Smart Objects > Convert to Smart Object) helps you resize without degrading it.
• Mask and Blend: Use layer masks to refine edges and blend with existing layers.
• Adjust Colors and Tones: Apply adjustment layers (e.g., Curves, Color Balance) to match the generated asset to your scene’s mood.
• Apply Filters: Use traditional Photoshop filters or Neural Filters such as Harmonization to further integrate the element.

This turns the AI from a simple generator into an assistant that provides raw material for your final artistic vision.

Practical Section: Enhancing an Image with AI in Photoshop

Here’s how these AI tools fit into a common workflow: enhancing a standard landscape photo.

1. Open Your Image: Launch Photoshop and choose File > Open to open your landscape photo.
2. Expand the Canvas with Generative Expand: Select the Crop tool from the Tools panel and drag the crop handles outward to give the image more room. Select “Generative Expand” in the Contextual Task Bar and choose a Firefly model in the model picker. Leave the prompt blank to let Photoshop extend the scene, or type something like “more sky and clouds.” Select Generate, then choose the best variation in the Properties panel.
3. Remove an Unwanted Object with Generative Fill: Notice a distracting power line or person in the background? Select it with a selection tool such as the Lasso Tool or Object Selection Tool. Select “Generative Fill” in the Contextual Task Bar, leave the prompt blank, and select Generate. Review the variations in the Properties panel and keep the best one.
4. Adjust the Mood with Neural Filters: Go to Filter > Neural Filters and turn on Landscape Mixer (Beta). Use its presets or season and time-of-day sliders, such as Sunset, to change the mood. Set the output to a new layer or Smart Filter so the edit stays non-destructive.
5. Refine and Export: Make final adjustments with traditional tools like Curves or Vibrance. Use layer masks on your generated layers to fine-tune blending. When you’re happy, save your work and export it in your chosen format.

This workflow shows how AI can turn what used to be hours of meticulous work into a few clicks and prompts, while you keep full creative control.

What to Watch Out For

Photoshop’s AI is powerful, but it has quirks and pitfalls that can derail an otherwise good workflow.

First, over-reliance on AI can lead to generic or “AI-looking” results. The best outcomes come from human-AI collaboration, with your artistic eye guiding the output. Don’t be afraid to refine masks or blend generated content by hand. Second, watch out for AI artifacts: subtle distortions, repeated patterns, or illogical elements, especially with complex prompts or unusual selections. Always zoom in and inspect generated areas closely. Third, check the system requirements. Adobe publishes minimum and recommended specs for Photoshop on its technical requirements page, and Beta Neural Filters may give variable results. Fourth, keep an eye on generative credits: generative features draw on your plan’s monthly allocation. Finally, be mindful of copyright and usage rights. Adobe states Firefly is designed to be commercially safe, but check the current terms, especially if you combine Firefly output with partner models or other AI tools.

Bottom Line

Tools like Generative Fill, Generative Expand, and Neural Filters have become a core part of a modern Photoshop workflow. They help artists and designers work faster, experiment more freely, and tackle edits that used to be prohibitively time-consuming. Understanding how to use AI in Photoshop is quickly becoming a baseline skill for anyone serious about image editing.

Our recommendation is simple: dive in and experiment. Treat AI as an intelligent assistant that amplifies your skills rather than a replacement for them. Start using these tools in your everyday projects, practise your prompting, and lean on Adobe’s help pages for the exact steps as features evolve.

FAQ

Is Photoshop’s AI free to use?

Photoshop requires an Adobe subscription. Generative features such as Generative Fill and Generative Expand use generative credits. Adobe says Creative Cloud plans include a monthly allocation of credits, that the number depends on your plan and the feature you use, and that you can buy more if you need them (Adobe: Generative credits overview).

What’s the difference between Generative Fill and Generative Expand?

Generative Fill adds, replaces or removes content inside a selection you make on the image (Adobe). Generative Expand extends the image beyond its original borders: you drag the Crop tool’s handles outward, then select Generative Expand in the Contextual Task Bar (Adobe).

Can I use Photoshop AI for commercial projects?

Adobe states that for Firefly features without the beta label, you can use Firefly-generated outputs in commercial projects, and that beta-feature outputs can also be used commercially unless the product explicitly says otherwise. Adobe also says it designed Firefly to be commercially safe by training on licensed and public domain content (Adobe Firefly FAQ). Because Generative Fill can also use partner models, and because the law is evolving, check Adobe’s current terms for the model you used.

What are the system requirements for Photoshop?

For Photoshop version 27.x and later on Windows, Adobe lists a multicore Intel, AMD or WinARM processor with AVX2 support, 8 GB of RAM (16 GB or more recommended), a GPU with DirectX 12 support (feature level 12_0 or later) and 1.5 GB of GPU memory (4 GB for 4K displays and greater), and 10 GB of available disk space (100 GB recommended). macOS requirements are listed separately. See Adobe’s full, current list on its Photoshop technical requirements page.

Do Photoshop’s AI features need an internet connection?

Firefly-powered features do: Adobe lists a stable internet connection among the requirements for using Firefly features (Adobe Firefly FAQ). Adobe also notes that some Neural Filters need internet connectivity for cloud processing (Adobe: Neural Filters gallery).

Related Reading

• Best AI Image Generator in 2026: Top Picks Reviewed
• Adobe Background Remover: Complete Guide for 2026
• Best AI Photo Editors in 2026: Expert Picks
• How to Write Better AI Image Prompts (2026 Guide)
• AI Video Editing

Sources

• Adobe: Explore beyond the canvas with Generative Expand (last updated Feb 23, 2026)
• Adobe: Edit images with Generative Fill (last updated Sep 15, 2026)
• Adobe: Photoshop on desktop technical requirements (last updated Sep 9, 2026)
• Adobe: Neural Filters gallery (last updated Oct 4, 2026)
• Adobe: Use Neural Filters to enhance images (last updated Feb 23, 2026)
• Adobe: Photoshop Neural Filters (Landscape Mixer season and Sunset sliders)
• Adobe Firefly FAQ (last updated May 13, 2026)
• Adobe: Generative credits overview (last updated Aug 17, 2026)